Question Details

(solution) Journal of Experimental Psychology: Human Learning and


Here attached is an article on Arthur Reber's 1976 experiment. Can you please explain the theory behind the experiment, the hypothesis (including independent and dependent variable) (the rationale behind the experiment), the procedure, what information did Reber found, the results and why did it happen?


Journal of Experimental Psychology:

 

Human Learning and Memory

 

1976, Vol. 2, No. 1, 88-94 Implicit Learning of Synthetic Languages:

 

The Role of Instructional Set

 

Arthur S. Reber

 

Brooklyn College

 

The effect of instructional set on implicit learning of a synthetic language

 

was explored. Specifically, the neutral, implicit instructions used in previous studies were compared with explicit instructions which directed subjects to search for the complex rules that determined letter orderings. The

 

subjects given the explicit instructions were poorer at memorizing exemplars

 

from the language, learned less about the underlying structure, and tended

 

to invent nonrepresentational rules. The results have strong implications

 

for a theory of implicit learning which stresses a nonconscious abstraction

 

system that operates when the stimulus environment exhibits exceedingly

 

complex structure, and subjects are not actively trying to break the code. Elsewhere implicit learning has been characterized as a process whereby a subject becomes sensitive to the structure inherent in

 

in a complex array by developing (implicitly) a conceptual model which reflects

 

the structure to some degree (Reber, 1967,

 

1969; Reber & Millward, 1968). It has

 

been a working hypothesis that the learning

 

process is fundamentally an abstraction of

 

information from the environment by the

 

subject without recourse to explicit strategies for responding or explicit systems for

 

encoding the stimuli. Support for the hypothesis has been indirect, consisting essentially of the failure of several groups of

 

researchers to find evidence of verbalizable,

 

explicit hypothesis testing or rule formation

 

even when subjects display highly efficient

 

exploitation of the patterns present in the

 

stimuli (Braine, 1963; Foss, 1968; Reber,

 

1967, 1969; Smith & Braine, in press).

 

Given the highly abstract nature of the

 

implicit learning process the evidence will

 

necessarily continue to be indirect. However, such evidence has a way of accumulating until a threshold of legitimacy is

 

achieved for the hypothesized abstraction.

 

This research was supported in part by grant

 

MH-20239-01 from the National Institute of Mental Health.

 

Requests for reprints should be sent to Arthur S.

 

Reber, Department of Psychology, Brooklyn College of CUNY, Brooklyn, New York 11210. This study provides an additional quantum

 

of data.

 

One of the obligations incurred whenever

 

a process is hypotheized is the specification

 

of the boundary conditions, the circumstances under which the presumed process

 

will occur. One limiting condition in implicit learning clearly must be the lower

 

bound on the complexity of the underlying

 

structure that defines the stimuli. That is,

 

the stimulus patterns must be such that they

 

bias against the possibility of there being

 

appropriate coding schemes available to the

 

subjects. The issue is essentially one of definition : If the subject can discover and

 

formalize the rule system that characterizes

 

the stimulus array then the experiment is

 

no longer an experiment in implicit learning; it becomes one in inductive rule learning or rule identification. It was this consideration that prompted the highly complex

 

structures in the past and dictates the similarly complex system in this study.

 

What is not so obvious, however, is the

 

question of the "set" of the subject when

 

confronted with these complex arrays. In

 

previous work great care has been taken to

 

ensure that subjects were neutral with regard to underlying structure, and thus presumably neutral with respect to the use of

 

explicit hypothesis-testing strategies. This

 

study was designed to explore the impact

 

that the operating set has upon the subject's

 

performance. The procedure used was IMPLICIT LEARNING OF SYNTHETIC LANGUAGE straightforward: The behavior of subjects

 

run under neutral instructions was compared with that of subjects who were given

 

general information about synthetic grammars and encouraged to undertake an explicit search for rules. More specifically,

 

the aim was to explore the differences between (a) subjects who maintained a relatively naive stance with regard to rule structure and operated in a neutral mode insofar

 

as the formulation of hypotheses and strategies is concerned, and (b) subjects who

 

actively searched for rules and operated in

 

an explicit hypothesis-testing mode.

 

METHOD

 

Stimulus Items

 

As in previous studies the stimulus items were

 

strings of letters generated by a finite-state grammar (Figure 1). This grammar may be characterized as a Markovian process in which each

 

permissible transition from any one state, Si, to

 

another state, Sj, generates a symbol. A grammatical string in the language is defined as any

 

sequence of permissible transitions leading from

 

the initial state, So, to the terminal state,So'. The

 

language is defined as all possible paths through

 

the system. For example, the sequence of states

 

So-Sa-Sa-Si-Sz-Ss-Si-So' generates the acceptable

 

sequence VXVPXVS. This particular grammar

 

generates exactly 43 permissible letter sequences

 

of lengths three through eight which were used as

 

the grammatical stimulus materials for the experiment. (See Chomsky & Miller, 1958, and Reber,

 

1967, for details of the procedures involved in these

 

calculations.) Subjects

 

The subjects were 20 undergraduates who served

 

as part of a course requirement. They were randomly assigned to the two groups. Procedure

 

The experiment was run in two parts, a learning phase and a testing phase. Prior to the beginning of the learning phase, subjects were read

 

the instructions appropriate for the group they

 

were in. Except for these instructions all subjects were run identically.

 

The instructions for the implicit group (Group

 

I) were as follows:

 

This is a. simple memory experiment. You will

 

see items made up of the letters PSTVX. They

 

will run from three to eight letters in length

 

and will be shown to you in groups of three

 

items each. After seeing each set of three items

 

I will give you a card and your task will be to 89 FIGURE 1. Schematic diagram of the finite-state

 

grammar used to generate the stimuli. (So = initial

 

state; So' = terminal state. The language is all

 

possible paths through the system.)

 

try to reproduce all three items. I will tell

 

you which ones you reproduced correctly. After

 

you have reproduced all three correctly two

 

times in a row we will go on to a new set of

 

three items.

 

The instructions for the explicit group (Group

 

E) were the same as above with the addition of

 

the following:

 

The order of letters in each item is determined

 

by a rather complex set of rules. The rules only

 

allow certain letters to follow other letters. Since

 

the task involves memorization of a large number of these complex strings of letters, it will be

 

to your advantage if you can figure out what the

 

rules are, which letters may follow other letters

 

and which ones may not. Such knowledge will

 

certainly help you in learning and memorizing

 

the items.

 

Learning. Fifteen of the 43 grammatical strings

 

were selected as stimulus items for this phase of the

 

experiment. They were selected as representative

 

of the possible types of grammatical strings; for

 

example, for each length, strings beginning with

 

both T and V were included, and for all lengths

 

where it was possible, an example of each of the

 

loops of the grammar (P, X, VPX) was used.

 

These 15 items were presented to subjects in five

 

sets of 3 items each. Each item was printed on a

 

separate card and presented through a viewing window for 5 sec. After the 3 items of a given set

 

were shown, the subject was given a card and

 

asked to reproduce all 3 items. There were no

 

time restrictions on the subjects although long response times were rare. Subjects were told which

 

items were correctly reproduced but no information

 

was given about the nature of the errors. Each

 

set was run repeatedly in the same order until the

 

criterion of two consecutive correct reproductions

 

was reached, after which the next set was presented. The order of presentation was varied randomly for each subject. All subjects continued

 

until all 15 items were learned. A S-min rest ARTHUR S. REBER 90 TABLE 1

 

MEAN ERRORS TO CRITERION BY GROUP

 

Set

 

M Group I

 

E 8.3

 

17.4 4.9

 

9.5 4.0

 

7.7 2.0

 

5.3 2.3

 

4.5 4.03

 

8.88 Note. Abbreviations: I = implicit instructions; E = explicit

 

instructions. period was allowed before the beginning of the

 

test phase.

 

Testing. Twenty-two of the remaining grammatical items were selected along with 22 items

 

that violated the rules of the grammar. Four of

 

these nongrammatical items were formed randomly

 

and contained multiple violations; the remaining

 

18 contained only single-letter violations.

 

The stimulus items were presented through the

 

same viewing window. The subjects' task was to

 

make a decision about the correctness or grammaticality of each item based upon what they had

 

learned during the initial memorization phase and

 

to then press one of two buttons marked "yes"

 

and "no." Note that none of the subjects had been

 

told at the outset that there would be a testing

 

phase, and that for the subjects in Group I this

 

was the first time that any reference to rules had

 

been made.

 

The list of 44 test items was presented twice,

 

making a total of 88 for each subject. All subjects were informed about the equal proportions of

 

grammatical and nongrammatical items. There was

 

no time limit on the subjects during this phase

 

although they were told that latencies were being

 

recorded. No feedback about the correctness of a

 

decision was given until the full set of 88 items

 

was completed. RESULTS1 is high. Overall, Type 1 strings 2 were

 

somewhat easier to learn than other types,

 

and this tendency was observed in both

 

groups. The other four types were equally

 

difficult although Group E subjects experienced uniformly more difficulty than Group

 

I subjects and made more error on all types.

 

Testing

 

Note first that the lack of feedback about

 

the correctness of a response served to keep

 

subjects at a constant level of performance

 

throughout the testing phase. Also, as in

 

previous work, subjects were generally not

 

aware of the fact that each test item was

 

presented twice. Thus in the following

 

analyses the full set of 88 test items is

 

treated as a single block. The data are

 

presented in Table 2.

 

TABLE 2

 

FREQUENCY OF G AND NG RESPONSES TO G

 

AND NG ITEMS IN BOTH GROUPS

 

Item

 

Response NG Total Group I

 

G

 

NG 334

 

106 99

 

341 433

 

447 Group E

 

G

 

NG 263

 

177 131

 

309 394

 

486 Learning Note. Group differences are highly significant, p < .001; correct responses for both groups are significantly better than

 

chance, fa < .001. Abbreviations; G = grammatical; NG

 

= nongrammatical; I = implicit instructions; E = explicit

 

instructions. The learning phase data are most easily

 

expressed in terms of errors to criterion. As

 

Table 1 shows there was a strong difference

 

between the groups with those given explicit

 

instructions performing significantly poorer,

 

*(18) = 4.75, SEAltt = 2.14. Trials to criterion data were comparable and are not

 

presented.

 

In general the learning data were similar

 

to those found in earlier work (Reber, 1967,

 

1969). Subjects in both groups eventually

 

adopted the procedure of focusing upon only

 

one or two strings per memorization trial,

 

a common strategy when information load 1

 

The rejection region throughout is p < .01

 

unless otherwise noted.

 

2

 

The finite-state grammar in Figure 1 generates

 

five basic sentence types. Each type is defined by

 

a path through the system with obligatory and

 

optional transitions, the latter being the loops

 

or recursions. The five types are as follows:

 

(1) T[P]TS; (2) T[P]TX[X](VPX[X])VS;

 

(3) T[P]TX[X](VPX[X])VPS; (4) V[X]

 

(VPX[X])VS; and (5) V[X] (VPX[X])VPS.

 

Note that Types 2 and 3 and Types 4 and 5 are

 

very similar differing only in the obligatory P in

 

the next-to-last position in Types 3 and 5; Type 1

 

appears, superficially, to be considerably simpler

 

than the other four. IMPLICIT LEARNING OF SYNTHETIC LANGUAGE

 

Both groups showed the fruits of the

 

learning phase and were able to discriminate

 

grammaticality (G) from nongrammaticality

 

(NG) at far better than chance levels, fs(9)

 

= 22.3 and 15.2 for Groups I and E, respectively, SEM = .59 and 1.54, respectively.

 

Moreover, every one of the 20 subjects

 

showed discriminability above chance.

 

_ However, the two groups differed from each

 

other considerably in this ability, with Group

 

E subjects being poorer than those given

 

the implicit instructions, <(18) = 6.24, SEAUt

 

= 1.74. Group I performance agreed nicely

 

with earlier findings and, indeed, is actually

 

a replication of work reported in Reber

 

(1967); Group E performance fell far below this level. Note also that Group E

 

shows a strong bias toward nongrammatical

 

responses. The argument is made later in

 

this article that this bias can be expected

 

under the assumption that the primary effect

 

of the specific instructions to these subjects

 

was to produce a tendency for them to develop rule systems which were not representative of the underlying structure.

 

A variety of other, more fine-grain,

 

analyses were carried out, none of which

 

provided any important insight into the

 

qualitative differences between the two

 

groups, and all of which were comparable to

 

similar results found in Reber (1967). For

 

example, the nongrammatical items with the

 

violation in either the initial or the terminal

 

letter were detected better than items with

 

the violation in internal positions. Items

 

which contained multiple violations were

 

easier to detect as nongrammatical than

 

items with single letter violations. The five

 

item types were all equally likely to be recognized as grammatical. Naturally, the overall level of performance in all these cases

 

was lower for Group E subjects but the

 

general pattern was the same in both groups.

 

Further, the latency data failed to reveal

 

group differences. Both groups had shorter

 

latency distributions for items on which

 

grammaticality was correctly assessed than

 

on items where an error was made, p < .01

 

for Group I and p < .02 for Group E (Kolmogorov-Smirnov test). There were no differences in response time to assign grammaticality as opposed to nongrammaticality 91 for either group, and there were no betweengroups differences.

 

The one statistic that does reveal an important quantitative difference between the

 

mode of operation of the two groups is the

 

probability of a subject making an error on

 

both presentations of a given item (Pe,e)

 

compared with the probability of an error

 

on only one of the two presentations (Pe,c

 

and Pc,e). In terms of a simple detection

 

model, Pc,c = fc + (1-*) g2

 

Pc,e = P e , c = ( l - £ ) < 7 ( l - 0 ) P... = (!-*) ( l - < 7 ) 2 , where k is a parameter which reflects the

 

subjects' level of apprehension of the grammatical relations, that is, the probability of

 

knowing the grammaticality of any given

 

item. The probability of a correct guess

 

is represented as g, and by virtue of the

 

equal proportions of G and NG items, g =

 

.50. The known value of P<.tC was used to

 

estimate k and thus "predict" the other

 

values.

 

In principle, k can be estimated from any

 

of the equations, although Pc,c is the appropriate source since this value alone is

 

assumed to contain instances where the subjects knew the grammatical status of the

 

letter strings. Further, estimating k in this

 

manner provides for a cleaner test of the

 

strong prediction of the model, that is,

 

Pc,e = Pe,e = Pe,e> which IS 3 direct result of fixing g at .50.

 

Interpreting the model is straightforward:

 

It is deemed appropriate only under the

 

condition where the subjects' decisions are

 

based upon an accurate (although partial)

 

representation of the rules of the grammar.

 

Inappropriate representative rules will create

 

instances where subjects incorrectly assume

 

that they know the grammatical status of

 

test items and will produce an inflated value

 

of Pe,e relative to Pc,e and Pe,c. Thus the

 

model serves as a sensitive test of the implicit nature of the subjects' behavior as it

 

pertains to the error data.

 

In evaluating the model x2 tests for goodness-of-fit were carried out with the values

 

of PeiC and Pc,e pooled. The pooling was

 

done so that any perturbations produced by 92 ARTHUR S. REBER

 

TABLE 3

 

TEST OF THE SIMPLE DETECTION MODEL

 

Probability

 

Result Group I

 

Predicted

 

Obtained

 

Group E

 

Predicted

 

Obtained .66

 

.66 .11

 

.10 .11

 

.11 .11

 

.13 .53

 

.53 .16

 

.12 .16

 

.12 .16

 

.23 vaue o

 

= .55 and .37 for Groups

 

I and E, respectively.

 

model is rejected for

 

Group E. x!(10) = 44.06, p < .001, bu

 

not for Group I, x2(10) = 8.86. fluctuations in these values would not contribute to x2, and thus any rejection of the

 

model must be due to the inflated value

 

of Pe,e- These fluctuations were, of course,

 

nonsystematic, as can be seen from the

 

group averages given in Table 3 where Pe,c

 

and Pc,e are essentially identical for both

 

groups.

 

Each individual subject was tested against

 

the model and the individual x2 values were

 

summed to produce the group results shown

 

in Table 3. Using p < .05 as the critical

 

value, 7 of the 10 subjects in Group E had

 

values of Pe,e in excess of what would be

 

expected by chance alone, while only 1 such

 

subject was found in Group I. For completeness, all predicted and obtained values

 

are given in Table 3, although the x2 tests

 

were all run with the pooling described

 

before.

 

The overall effects are quite clear. The

 

model is clearly an inappropriate characterization of the behavior of Group E subjects

 

while for Group I it is well within the

 

expected range. Subjects who operate in

 

an implicit mode develop abstract representations which are accurate (if partial) reflections of the structure of the stimulus

 

items. Subjects who are instructed to perform explicit rule searches develop abstractions which also contain representations

 

that are inaccurate reflections of the underlying structure. It is also worth noting that,

 

independent of the appropriateness of the

 

rule systems developed, both groups are

 

equally consistent in applications. Summing the values of .Pc,c and Pe,e yields .79 for

 

Group I and .76 for Group E, so that in

 

some sense each group has learned the same

 

"number" of rules.

 

Careful subject-by-subject analyses were

 

carried out in an effort to pinpoint qualitative

 

differences, particularly on those items responded to consistently. Occasionally systematicity was discovered, such as one E

 

subject who (erroneously) rejected any

 

item with a letter repeated more than four

 

times (despite the fact that TPPPPPTS

 

was among the set of learning items), and

 

another who accepted (also erroneously)

 

any item where the P-cycle was misplaced

 

(e.g., TTPPPXVS). These cases, however, were relatively rare and account for

 

a trivial amount of the large group differences that were found. The rich and

 

complex rule system which, as is argued

 

later, is necessary to produce implict learning, carries with it the serious liability that

 

subjects' abstractions will be similarly rich,

 

complex, and intractable.

 

DISCUSSION

 

There are several straightforward conclusions to be drawn from these data: (a) Subjects who engaged in an explicit search for

 

rules that define a complex structure performed more poorly in memorizing exemplars of the structure than subjects who

 

operated in a more neutral, implicit fashion,

 

(b) Although taken to the same learning

 

criterion, subjects operating in the explicit

 

mode ultimately learned less about the

 

underlying properties of the complex stimuli

 

than subjects in the implicit mode, (c) Explicit search for rules produced a strong

 

tendency for subjects to induce or invent

 

rules which were not accurate representations of the complex stimulus structure; this

 

tendency was not observed in subjects given

 

the implicit instructions.

 

The critical word in each of the foregoing

 

conclusions is complex. Except for it, these

 

conclusions would be at variance with the

 

rule-learning literature in which instructions that focus the subject's attention upon

 

the rule system generally accelerate learning.

 

In these other typically explicit rule-learning

 

experiments, since the stimulus patterns are IMPLICIT LEARNING OF SYNTHETIC LANGUAGE relatively simple and codable, a subject with

 

a reasonably rich stock of heuristics and

 

problem-solving strategies is going to find

 

their implementation rewarded. For example, essentially all of the work in serial

 

pattern learning (see Jones, 1974) has been

 

carried out using stimulus sequences whose

 

underlying structures can be coded by a subject equipped with search strategies based

 

upon devices like alternating events, eventrun length, complementation, and so forth.

 

However powerful these strategies may

 

be, the structure of the strings of letters

 

generated by the grammar in Figure 1 is

 

simply too rich to be coded by a subject

 

using them in the short time allowed. The

 

implication is that the explicit instructions

 

disrupted performance by inviting, indeed

 

encouraging, subjects to engage in futile

 

rule-search procedures and to elaborate rule

 

systems which were frequently nonrepresentational. The slower learning rate, the bias

 

toward assigning nongrammaticality, and the

 

inflated value of Pe,e all support this interpretation. It is important to note that the

 

instructions to Group E are not specifically

 

misleading. They interfere with performance, not because they mislead the subjects, but because they put them in an

 

operating set where they mislead themselves.

 

For example, one not atypical subject exhibited consternation during the postexperimental debriefing when she discovered that

 

her elaborate and sophisticated efforts to

 

find remote, deterministic contingencies between letters were doomed to fail. It is

 

trivially true, then, that searching for rules

 

will not work unless you can find them.

 

It should be emphasized here that explicit

 

rule search can jeopardize its user in

 

another, perhaps more significant, way. In

 

addition to producing poor performance because of a failure to find well-formed rules,

 

engaging in explicit rule search acts to mask

 

the implicit learning process. As has been

 

suggested elsewhere (Reber, 1967; Reber,

 

& Millward, 1968), the implicit acquisition

 

process seems to be most effective when the

 

subjects are in a relatively neutral, passive

 

set and allow themselves to be inundated by

 

the stimulus materials. The efforts on the

 

part of Group E subjects to break the code 93. precludes the operation of this implicit mode.

 

A serious difficulty with these experiments

 

on implicit learning is determining just how

 

they blend in with the traditional work on

 

rule learning and rule identification. Although this study is indeed an investigation of the conditions under which subjects

 

acquire rule-governed behaviors, it seems

 

prudent at this stage to allow the term

 

implicit learning to maintain definitional

 

integrity apart from both rule learning and

 

rule identification, as those terms are traditionally used. The separation from rule

 

identification seems straightforward; the

 

process here is basically one of systematic

 

testing of existing hypotheses and rules. It

 

is an interesting problem but one very different from inquiring how those rules came

 

to be.

 

The argument for the nonsynonymity of

 

implicit learning and rule learning is subtler.

 

As a first approximation it is proposed that

 

rule learning subsumes at least two elementary processes: a primitive process of

 

apprehending structure by attending to frequency cues, and a more explicit process

 

whereby various mnemonics, heuristics, and

 

strategies are engaged to induce a representational system. The former is what is

 

defined here as implicit learning, and it has

 

certain conceptual similarities with the

 

differentiation

 

component of perceptual

 

learning (see Gibson, 1969). The latter

 

is what is listed in the psychological lexicon

 

under rule learning.

 

The paradigmatic confusion, however, is

 

not necessarily lessened by this classification. .In practice it is exceedingly difficult

 

to identify the boundary between rule learning and rule identification. I know of no

 

published report on rule learning in adults

 

where the rules to be learned were not immediate or trivial generalizations of well-practiced and readily retrievable rule systems.

 

Despite the nomenclature, cognitive psychologists rarely study rule learning.

 

The work of Miller and his associates on

 

artificial-grammar learning (Miller, 1967,

 

chap. 7) is illustrative. Their work, which

 

was begun using grammars of a complexity

 

approaching that of Figure 1 (Miller, 1958;

 

Shipstone, 1960), was shifted over to rela- ARTHUR S. REBER

 

tively simple systems on the grounds that

 

the synthetic languages used initially were

 

too intricate for "an afternoon in the laboratory." The outcome of "Project Grammarama" thus became, as Miller recognized,

 

an interesting procedure for evaluating the

 

process of explicit-rule induction. Essentially nothing was learned about the implicit

 

acquisition of highly complex systems like

 

languages. The problem was th...

 


Solution details:

Pay using PayPal (No PayPal account Required) or your credit card . All your purchases are securely protected by .
SiteLock

About this Question

STATUS

Answered

QUALITY

Approved

DATE ANSWERED

Sep 13, 2020

EXPERT

Tutor

ANSWER RATING

GET INSTANT HELP/h4>

We have top-notch tutors who can do your essay/homework for you at a reasonable cost and then you can simply use that essay as a template to build your own arguments.

You can also use these solutions:

  • As a reference for in-depth understanding of the subject.
  • As a source of ideas / reasoning for your own research (if properly referenced)
  • For editing and paraphrasing (check your institution's definition of plagiarism and recommended paraphrase).
This we believe is a better way of understanding a problem and makes use of the efficiency of time of the student.

NEW ASSIGNMENT HELP?

Order New Solution. Quick Turnaround

Click on the button below in order to Order for a New, Original and High-Quality Essay Solutions. New orders are original solutions and precise to your writing instruction requirements. Place a New Order using the button below.

WE GUARANTEE, THAT YOUR PAPER WILL BE WRITTEN FROM SCRATCH AND WITHIN A DEADLINE.

Order Now